Signal complexity and synchrony of epileptic seizures: is there an identifiable preictal period?
Introduction
Epileptic seizures are characterized by rapidly evolving dynamic changes that are in part reflected in the EEG. The hypothesis that epileptic seizures can be predicted by sophisticated methods of signal analysis is predicated on the existence of a preictal period and the ability to detect or reveal these periods using these methods.
There recently has been increased interest in examining ictal and interictal EEG changes using new time–frequency analyses and various non-linear dynamic approaches (see partial review in Le Van Quyen et al., 2001). Previous applications of the matching pursuit (Bergey and Franaszczuk, 2001) and Gabor atom density (GAD) (Jouny et al., 2003) methods have shown that ictal periods are consistently associated with changes in complexity of the EEG signal. Based on time–frequency decomposition, GAD measures the complexity of a signal as the number of elementary components needed to represent the EEG signal. The method appears to be very sensitive and specific for detecting intracranial ictal activity from a selected channel. The earliest changes during a seizure have been seen in the channels closest to the region of seizure onset.
Synchronized activity of neural populations also changes significantly during seizures. Measure S, based on the residual covariance matrix of a multivariable autoregressive model (MVAR) allows for a measurement of the degree of synchrony in the EEG signal. Previous applications of this measure to intracranial recordings have also revealed persistent changes in synchrony in the postictal period (Franaszczuk and Bergey, 1999).
In the studies reported here we have applied these methods to the data provided for the First International Consortium Workshop for Seizure Prediction. We tested the hypothesis that changes in signal complexity and synchrony precede the visually recognizable EEG changes or the clinical manifestations of the seizures. Specifically we will be determining whether a preictal period exists, distinct from the interictal period. If indeed this is discernable then this may facilitate seizure prediction. Since epileptic seizures often cluster, particularly in the environment of epilepsy monitoring units, the time between seizures may have a significant influence on the observations during the interictal period.
Section snippets
Data sets
The description of the data set can be found in the introductory article (Lehnertz and Litt, 2005). We present here only analyses for data sets B and E, which included long continuous multi-day recordings with ictal and postictal periods. Other sets were either discontinuous or did not include actual seizures.
Gabor atom density
The Gabor atom density (GAD) method is based on the matching pursuit (MP) algorithm (Mallat and Zhang, 1993) and is fully described elsewhere (Jouny et al., 2003). A short introduction to
Results
Fig. 1A shows the continuous plot of GADT for one channel (right amygdala) closest to the focus for the 3 last days of the data set E. Seizures appear clearly; they result in large increases of the GADT levels. Known seizures are marked with black triangles. The GADT level during reference period is 0.075+/−0.015. The mean level throughout the recording is 0.093+/−0.041. The mean maximum GADT value during seizures is 0.41+/−0.03 ranging from 0.36 to 0.44, an average of 300% increase compared to
Discussion
Recurrent seizures are comprised of ictal periods and interictal periods. The ictal period is conventionally described by the visually apparent changes in the EEG consistent with seizure activity. This period includes the period of clinical manifestations, but may include initial EEG changes that are visually detectable, but may not have accompanying clinical manifestations. The interictal period includes the postictal period and the preictal period and additional background activity, dependent
Acknowledgements
This research was supported partially by NIH grant NS 33732.
References (15)
- et al.
Epileptic seizures are characterized by changing signal complexity
Clin Neurophysiol
(2001) - et al.
Regional coherence and the transfer of ictal activity during seizure onset in the medial temporal lobe
Electroencephalogr Clin Neurophysiol
(1992) - et al.
Analysis of mesial temporal seizure onset and propagation using the directed transfer function method
Electroenceph Clin Neurophysiol
(1994) - et al.
Amygdala-hippocampus relationships in temporal lobe seizures: a phase-coherence study
Epilepsy Res
(1996) - et al.
Characterization of epileptic seizure dynamics using Gabor atom density
Clin Neurophysiol
(2003) - et al.
Does interictal spiking change prior to seizures?
Electroencephalogr Clin Neurophysiol
(1991) Non-linear time series analysis of intracranial EEG recordings in patients with epilepsy—an overview
Int J Psychophysiol
(1999)
Cited by (53)
Application of Higuchi's fractal dimension from basic to clinical neurophysiology: A review
2016, Computer Methods and Programs in BiomedicineResponsive neurostimulation for the treatment of medically intractable epilepsy
2013, Brain Research BulletinSeizure prediction: Methods
2011, Epilepsy and BehaviorSeizure Prediction and its Applications
2011, Neurosurgery Clinics of North AmericaPatient-specific bivariate-synchrony-based seizure prediction for short prediction horizons
2010, Epilepsy Research